Retrieval of super‐virtual refraction by cross‐correlation

Topography and severe variations of near‐surface layers lead to travel‐time perturbations for the events in seismic exploration. Usually, these perturbations could be estimated and eliminated by refraction technology. The virtual refraction method is a relatively new technique for retrieval of refraction information from seismic records contaminated by noise. Based on the virtual refraction, this paper proposes super‐virtual refraction interferometry by cross‐correlation to retrieve refraction wavefields by summing the cross‐correlation of raw refraction wavefields and virtual refraction wavefields over all receivers located outside the retrieved source and receiver pair. This method can enhance refraction signal gradually as the source–receiver offset decreases. For further enhancement of refracted waves, a scheme of hybrid virtual refraction wavefields is applied by stacking of correlation‐type and convolution‐type super‐virtual refractions. Our new method does not need any information about the near‐surface velocity model, which can solve the problem of directly unmeasured virtual refraction energy from the virtual source at the surface, and extend the acquisition aperture to its maximum extent in raw seismic records. It can also reduce random noise influence in raw seismic records effectively and improve refracted waves’ signal‐to‐noise ratio by a factor proportional to the square root of the number of receivers positioned at stationary‐phase points, based on the improvement of virtual refraction's signal‐to‐noise ratio. Using results from synthetic and field data, we show that our new method is effective to retrieve refraction information from raw seismic records and improve the accuracy of first‐arrival picks.     

Automatic stacking‐velocity estimation using similarity‐weighted clustering

Local seismic event slopes contain subsurface velocity information and can be used to estimate seismic stacking velocity. In this paper, we propose a novel approach to estimate the stacking velocity automatically from seismic reflection data using similarity‐weighted k‐means clustering, in which the weights are local similarity between each trace in common midpoint gather and a reference trace. Local similarity reflects the local signal‐to‐noise ratio in common midpoint gather. We select the data points with high signal‐to‐noise ratio to be used in the velocity estimation with large weights in mapped traveltime and velocity domain by similarity‐weighted k‐means clustering with thresholding. By using weighted k‐means clustering, we make clustering centroids closer to those data points with large weights, which are more reliable and have higher signal‐to‐noise ratio. The interpolation is used to obtain the whole velocity volume after we have got velocity points calculated by weighted k‐means clustering. Using the proposed method, one obtains a more accurate estimate of the stacking velocity because the similarity‐based weighting in clustering takes into account the signal‐to‐noise ratio and reliability of different data points in mapped traveltime and velocity domain. In order to demonstrate that, we apply the proposed method to synthetic and field data examples, and the resulting images are of higher quality when compared with the ones obtained using existing methods.     

The impact of surface‐wave attenuation on 3‐D seismic survey design

Three‐dimensional seismic survey design should provide an acquisition geometry that enables imaging and amplitude‐versus‐offset applications of target reflectors with sufficient data quality under given economical and operational constraints. However, in land or shallow‐water environments, surface waves are often dominant in the seismic data. The effectiveness of surface‐wave separation or attenuation significantly affects the quality of the final result. Therefore, the need for surface‐wave attenuation imposes additional constraints on the acquisition geometry. Recently, we have proposed a method for surface‐wave attenuation that can better deal with aliased seismic data than classic methods such as slowness/velocity‐based filtering. Here, we investigate how surface‐wave attenuation affects the selection of survey parameters and the resulting data quality. To quantify the latter, we introduce a measure that represents the estimated signal‐to‐noise ratio between the desired subsurface signal and the surface waves that are deemed to be noise. In a case study, we applied surface‐wave attenuation and signal‐to‐noise ratio estimation to several data sets with different survey parameters. The spatial sampling intervals of the basic subset are the survey parameters that affect the performance of surface‐wave attenuation methods the most. Finer spatial sampling will reduce aliasing and make surface‐wave attenuation easier, resulting in better data quality until no further improvement is obtained. We observed this behaviour as a main trend that levels off at increasingly denser sampling. With our method, this trend curve lies at a considerably higher signal‐to‐noise ratio than with a classic filtering method. This means that we can obtain a much better data quality for given survey effort or the same data quality as with a conventional method at a lower cost.     

CRS-stack-based seismic imaging considering top-surface topography

In this case study we consider the seismic processing of a challenging land data set from the Arabian Peninsula. It suffers from rough top‐surface topography, a strongly varying weathering layer, and complex near‐surface geology. We aim at establishing a new seismic imaging workflow, well‐suited to these specific problems of land data processing. This workflow is based on the common‐reflection‐surface stack for topography, a generalized high‐density velocity analysis and stacking process. It is applied in a non‐interactive manner and provides an entire set of physically interpretable stacking parameters that include and complement the conventional stacking velocity. The implementation introduced combines two different approaches to topography handling to minimize the computational effort: after initial values of the stacking parameters are determined for a smoothly curved floating datum using conventional elevation statics, the final stack and also the related residual static correction are applied to the original prestack data, considering the true source and receiver elevations without the assumption of nearly vertical rays. Finally, we extrapolate all results to a chosen planar reference level using the stacking parameters. This redatuming procedure removes the influence of the rough measurement surface and provides standardized input for interpretation, tomographic velocity model determination, and post‐stack depth migration. The methodology of the residual static correction employed and the details of its application to this data example are discussed in a separate paper in this issue. In view of the complex near‐surface conditions, the imaging workflow that is conducted, i.e. stack – residual static correction – redatuming – tomographic inversion – prestack and post‐stack depth migration, leads to a significant improvement in resolution, signal‐to‐noise ratio and reflector continuity.     

Arctangent function‐based third derivative attribute for characterisation of faults

Using seismic attributes such as coherence and curvature to characterise faults not only can improve the efficiency of seismic interpretation but also can expand the capability to detect faults. The coherence and curvature have been widely applied to characterising faults for years. These two methods detect faults based on the similarity of seismic waveforms and shapes of the reflectors, respectively, and they are complementary to each other and both have advantages and disadvantages in fault characterisation. A recent development in fault characterisation based on reflector shapes has been the use of the rate of change of curvature. Through an application to the seismic data from Western Tazhong of the Tarim Basin, China, it was demonstrated that the rate of change of curvature is more capable of detecting subtle faults having quite small throws and heaves. However, there often exist multiple extreme values indicating the same fault when applying the rate of change of curvature, which significantly degrades the signal‐to‐noise ratio of the computation result for multiple extrema interfering with each other. To resolve this problem, we propose the use of a linear combination of arctangent and proportional functions as the directrix of a cylindrical surface to fit the fault model and calculate its third derivative, which can then be used to characterise the fault. Through an application to the 3D seismic data from Western Tazhong of the Tarim Basin, the results show that the proposed method not only retains the same capability to detect subtle faults having small throws as the curvature change rate but also greatly improves the signal‐to‐noise ratio of the calculated result.     

Trans‐dimensional Bayesian inversion of controlled‐source electromagnetic data in the German North Sea

This paper presents the first controlled‐source electromagnetic survey carried out in the German North Sea with a recently developed seafloor‐towed electrical dipole–dipole system, i.e., HYDRA II. Controlled‐source electromagnetic data are measured, processed, and inverted in the time domain to estimate an electrical resistivity model of the sub‐seafloor. The controlled‐source electromagnetic survey targeted a shallow, phase‐reversed, seismic reflector, which potentially indicates free gas. To compare the resistivity model to reflection seismic data and draw a combined interpretation, we apply a trans‐dimensional Bayesian inversion that estimates model parameters and uncertainties, and samples probabilistically over the number of layers of the resistivity model. The controlled‐source electromagnetic data errors show time‐varying correlations, and we therefore apply a non‐Toeplitz data covariance matrix in the inversion that is estimated from residual analysis. The geological interpretation drawn from controlled‐source electromagnetic inversion results and borehole and reflection seismic data yield resistivities of ～1 Ωm at the seafloor, which are typical for fine‐grained marine deposits, whereas resistivities below ～20 mbsf increase to 2–4 Ωm and can be related to a transition from fine‐grained (Holocene age) to unsorted, coarse‐grained, and compacted glacial sediments (Pleistocene age). Interface depths from controlled‐source electromagnetic inversion generally match the seismic reflector related to the contrast between the different depositional environments. Resistivities decrease again at greater depths to ～1 Ωm with a minimum resistivity at ～300 mbsf where a seismic reflector (that marks a major flooding surface of late Miocene age) correlates with an increased gamma‐ray count, indicating an increased amount of fine‐grained sediments. We suggest that the grain size may have a major impact on the electrical resistivity of the sediment with lower resistivities for fine‐grained sediments. Concerning the phase‐reversed seismic reflector that was targeted by the survey, controlled‐source electromagnetic inversion results yield no indication for free gas below it as resistivities are generally elevated above the reflector. We suggest that the elevated resistivities are caused by an overall decrease in porosity in the glacial sediments and that the seismic reflector could be caused by an impedance contrast at a thin low‐velocity layer. Controlled‐source electromagnetic interface depths near the reflector are quite uncertain and variable. We conclude that the seismic interface cannot be resolved with the controlled‐source electromagnetic data, but the thickness of the corresponding resistive layer follows the trend of the reflector that is inclined towards the west.     

Burying receivers for improved time‐lapse seismic repeatability: CO2CRC Otway field experiment

4D seismic is widely used to remotely monitor fluid movement in subsurface reservoirs. This technique is especially effective offshore where high survey repeatability can be achieved. It comes as no surprise that the first 4D seismic that successfully monitored the CO₂ sequestration process was recorded offshore in the Sleipner field, North Sea. In the case of land projects, poor repeatability of the land seismic data due to low S/N ratio often obscures the time‐lapse seismic signal. Hence for a successful on shore monitoring program improving seismic repeatability is essential. Stage 2 of the CO2CRC Otway project involves an injection of a small amount (around 15,000 tonnes) of CO₂/CH₄ gas mixture into a saline aquifer at a depth of approximately 1.5 km. Previous studies at this site showed that seismic repeatability is relatively low due to variations in weather conditions, near surface geology and farming activities. In order to improve time‐lapse seismic monitoring capabilities, a permanent receiver array can be utilised to improve signal to noise ratio and hence repeatability. A small‐scale trial of such an array was conducted at the Otway site in June 2012. A set of 25 geophones was installed in 3 m deep boreholes in parallel to the same number of surface geophones. In addition, four geophones were placed into boreholes of 1–12 m depth. In order to assess the gain in the signal‐to‐noise ratio and repeatability, both active and passive seismic surveys were carried out. The surveys were conducted in relatively poor weather conditions, with rain, strong wind and thunderstorms. With such an amplified background noise level, we found that the noise level for buried geophones is on average 20 dB lower compared to the surface geophones. The levels of repeatability for borehole geophones estimated around direct wave, reflected wave and ground roll are twice as high as for the surface geophones. Both borehole and surface geophones produce the best repeatability in the 30–90 Hz frequency range. The influence of burying depth on S/N ratio and repeatability shows that significant improvement in repeatability can be reached at a depth of 3 m. The level of repeatability remains relatively constant between 3 and 12 m depths.     

Suppression of near‐surface scattered body‐to‐surface waves: a steerable and nonlinear filtering approach

We present an approach based on local‐slope estimation for the separation of scattered surface waves from reflected body waves. The direct and scattered surface waves contain a significant amount of seismic energy. They present great challenges in land seismic data acquisition and processing, particularly in arid regions with complex near‐surface heterogeneities (e.g., dry river beds, wadis/large escarpments, and karst features). The near‐surface scattered body‐to‐surface waves, which have comparable amplitudes to reflections, can mask the seismic reflections. These difficulties, added to large amplitude direct and back‐scattered surface (Rayleigh) waves, create a major reduction in signal‐to‐noise ratio and degrade the final sub‐surface image quality. Removal of these waves can be difficult using conventional filtering methods, such as an filter, without distorting the reflected signal. The filtering algorithm we present is based on predicting the spatially varying slope of the noise, using steerable filters, and separating the signal and noise components by applying a directional nonlinear filter oriented toward the noise direction to predict the noise and then subtract it from the data. The slope estimation step using steerable filters is very efficient. It requires only a linear combination of a set of basis filters at fixed orientation to synthesize an image filtered at an arbitrary orientation. We apply our filtering approach to simulated data as well as to seismic data recorded in the field to suppress the scattered surface waves from reflected body waves, and we demonstrate its superiority over conventional techniques in signal preservation and noise suppression.     

An automated cross‐correlation based event detection technique and its application to a surface passive data set

In studies on heavy oil, shale reservoirs, tight gas and enhanced geothermal systems, the use of surface passive seismic data to monitor induced microseismicity due to the fluid flow in the subsurface is becoming more common. However, in most studies passive seismic records contain days and months of data and manually analysing the data can be expensive and inaccurate. Moreover, in the presence of noise, detecting the arrival of weak microseismic events becomes challenging. Hence, the use of an automated, accurate and computationally fast technique for event detection in passive seismic data is essential. The conventional automatic event identification algorithm computes a running‐window energy ratio of the short‐term average to the long‐term average of the passive seismic data for each trace. We show that for the common case of a low signal‐to‐noise ratio in surface passive records, the conventional method is not sufficiently effective at event identification. Here, we extend the conventional algorithm by introducing a technique that is based on the cross‐correlation of the energy ratios computed by the conventional method. With our technique we can measure the similarities amongst the computed energy ratios at different traces. Our approach is successful at improving the detectability of events with a low signal‐to‐noise ratio that are not detectable with the conventional algorithm. Also, our algorithm has the advantage to identify if an event is common to all stations (a regional event) or to a limited number of stations (a local event). We provide examples of applying our technique to synthetic data and a field surface passive data set recorded at a geothermal site.     

在随机噪音背景下地震反射信号的增强

提高地震资料的信噪比，增强反射波同相轴连续性是地震勘探的基本问题之一。本文针对因存在噪音和信号比较弱，致使正常时差（NMO）校正加方法难以奏效的问题，提出提高信噪比和增强反射信号同相轴连续性的共反射面元（CRS）迭加方法，该方法将来自菲涅尔带的信息经校正迭加在一起，计算结果表明，该方法能有效提高信噪比，增强反射信号同相同的连续性。     

Non‐hyperbolic common reflection surface

The method of common reflection surface (CRS) extends conventional stacking of seismic traces over offset to multidimensional stacking over offset‐midpoint surfaces. We propose a new form of the stacking surface, derived from the analytical solution for reflection traveltime from a hyperbolic reflector. Both analytical comparisons and numerical tests show that the new approximation can be significantly more accurate than the conventional CRS approximation at large offsets or at large midpoint separations while using essentially the same parameters.     

The common focal point and common reflection surface methodologies: subsets or complements?

The common focal point (CFP) method and the common reflection surface (CRS) stack method are compared. The CRS method is a fast, highly automated procedure that provides high S/N ratio simulation of zero‐offset (ZO) images by combining, per image point, the reflection energy of an arc segment that is tangential to the reflector. It uses smooth parametrized two‐way stacking operators, based on a data‐driven triplet of attributes in 2D (eight parameters in 3D). As a spin‐off, the attributes can be used for several applications, such as the determination of the geometrical spreading factor, multiple prediction, and tomographic inversion into a smooth background velocity model. The CFP method aims at decomposing two‐way seismic reflection data into two full‐aperture one‐way propagation operators. By applying an iterative updating procedure in a half‐migrated domain, it provides non‐smooth focusing operators for prestack imaging using only the energy from one focal point at the reflector. The data‐driven operators inhibit all propagation effects of the overburden. The CFP method provides several spin‐offs, amongst which is the CFP matrix related to one focal point, which displays the reflection amplitudes as measured at the surface for each source–receiver pair. The CFP matrix can be used to determine the specular reflection source–receiver pairs and the Fresnel zone at the surface for reflection in one single focal point. Other spin‐offs are the prediction of internal multiples, the determination of reflectivity effects, velocity‐independent redatuming and tomographic inversion to obtain a velocity–depth model. The CFP method is less fast and less automated than the CRS method. From a pointwise comparison of features it is concluded that one method is not a subset of the other, but that both methods can be regarded as being to some extent complementary.     

汶川地震断裂带东北端浅部结构的人工地震探测

结合汶川地震断裂带动态监测,利用快速响应探测系统,开展了断层带浅部结构人工地震探测.针对地震断裂带动态监测条件下的复杂波场和低信噪比的情况,在f-k波场分离的基础上,分别利用了折射波共中心点成像、面波速度反演、反射波叠加成像方法,进行了浅层断层和构造成像处理,并对处理结果进行了综合解释,给出了断裂带浅部断层分布和速度特征.为汶川地震龙门山断裂带东北端动态监测提供了基础结构信息,所发展的断裂带快速响应探测技术对于地震应急动态监测具有重要意义.     

基于S变换的软阈值滤波在深地震反射数据处理中的应用

深地震反射原始单炮数据是非平稳的弱能量反射信号,信噪比较低.如何提高信噪比一直是深地震反射数据前处理中的一大难题.S变换是一种适用于分析非平稳信号的时频变换方法.同其他分析时变信号的方法相比,S变换的基本小波不必满足小波在时间域均值为零的容许性条件,它的时频分辨率与分析信号的频率有关,且其在时间域的积分可以得到傅里叶频谱,其反变换也简单.因此,S变换容易表示深地震反射信号复杂的时频特性.本文在S变换的基础上,利用软阈值滤波方法对深地震反射数据进行处理,实验结果表明,该方法有效地提高了信噪比,压制了有效频带范围内的混频干扰,突出了弱反射信号,使得波组信息更加丰富,有利于连续追踪有效反射波组和识别薄地层,特别是提高了深部Moho界面反射层位的分辨率,为深地震反射剖面后续处理和准确解释奠定了基础.     

Vertical seismic profiling using a daisy‐chained deployment of fibre‐optic cables in four wells simultaneously – Case study at the Ketzin carbon dioxide storage site

The geological storage of carbon dioxide is considered as one of the measures to reduce greenhouse gas emissions and to mitigate global warming. Operators of storage sites are required to demonstrate safe containment and stable behaviour of the storage complex that is achieved by geophysical and geochemical monitoring, combined with reservoir simulations. For site characterization, as well as for imaging the carbon dioxide plume in the reservoir complex and detecting potential leakage, surface and surface‐borehole time‐lapse seismic monitoring surveys are the most widespread and established tools. At the Ketzin pilot site for carbon dioxide storage, permanently installed fibre‐optic cables, initially deployed for distributed temperature sensing, were used as seismic receiver arrays, demonstrating their ability to provide high‐resolution images of the storage formation. A vertical seismic profiling experiment was acquired using 23 source point locations and the daisy‐chained deployment of a fibre‐optic cable in four wells as a receiver array. The data were used to generate a 3D vertical seismic profiling cube, complementing the large‐scale 3D surface seismic measurements by a high resolution image of the reservoir close to the injection well. Stacking long vibro‐sweeps at each source location resulted in vertical seismic profiling shot gathers characterized by a signal‐to‐noise ratio similar to gathers acquired using geophones. A detailed data analysis shows strong dependency of data quality on borehole conditions with significantly better signal‐to‐noise ratio in regions with good coupling conditions.     

Enhancement of Deep Seismic Reflections in Pre-stack Data by Adaptive Filtering

—Adaptive filters offer advantages over Wiener filters for time-varying processes. They are used for deconvolution of seismic data which exhibit non-stationary behavior, and seldom for noise reduction. Different algorithms for adaptive filtering exist. The least-mean-squares (LMS) algorithm, because of its simplicity, has been widely applied to data from different fields that fall outside geophysics. The application of the LMS algorithm to improve the signal-to-noise ratio in deep reflection seismic pre-stack data is studied in this paper. Synthetic data models and field data from the DEKORP project are used to this end.¶Three adaptive filter techniques, one-trace technique, two-trace technique and time-slice technique, are examined closely to establish the merits and demerits of each technique. The one-trace technique does not improve the signal-to-noise ratio in deep reflection seismic data where signal and noise cover the same frequency range. With the two-trace technique, the strongest noise reduction is achieved for small noise on the data. The filter efficiency decreases rapidly with increasing noise. Furthermore, the filter performance is poor upon application to common-midpoint (CMP) gathers with no normal-moveout (NMO) corrections. Application of the two-trace method to seismic traces before dynamic correction results in gaps in the signal along the reflection hyperbolas. The time-slice technique, introduced in this paper, offers the best answer. In this case, the one-trace technique is applied to the NMO-corrected gathers across all traces in each gather at each time to separate the low-wavenumber component of the signal in offset direction from the high-wavenumber noise component. The stacking velocities used for the dynamic correction do not need to be known very accurately because in deep reflection seismics, residual moveouts are small and have only a minor influence on the results of the adaptive time-slice technique. Noise reduction is more significant with the time-slice technique than with the two-trace technique. The superiority of the adaptive time-slice technique is demonstrated with the DEKORP data.     

Retrieving virtual reflection responses at drill‐bit positions using seismic interferometry with drill‐bit noise

In the field of seismic interferometry, researchers have retrieved surface waves and body waves by cross‐correlating recordings of uncorrelated noise sources to extract useful subsurface information. The retrieved wavefields in most applications are between receivers. When the positions of the noise sources are known, inter‐source interferometry can be applied to retrieve the wavefields between sources, thus turning sources into virtual receivers. Previous applications of this form of interferometry assume impulsive point sources or transient sources with similar signatures. We investigate the requirements of applying inter‐source seismic interferometry using non‐transient noise sources with known positions to retrieve reflection responses at those positions and show the results using synthetic drilling noise as source. We show that, if pilot signals (estimates of the drill‐bit signals) are not available, it is required that the drill‐bit signals are the same and that the phases of the virtual reflections at drill‐bit positions can be retrieved by deconvolution interferometry or by cross‐coherence interferometry. Further, for this case, classic interferometry by cross‐correlation can be used if the source power spectrum can be estimated. If pilot signals are available, virtual reflection responses can be obtained by first using standard seismic‐while‐drilling processing techniques such as pilot cross‐correlation and pilot deconvolution to remove the drill‐bit signatures in the data and then applying cross‐correlation interferometry. Therefore, provided that pilot signals are reliable, drill‐bit data can be redatumed from surface to borehole depths using this inter‐source interferometry approach without any velocity information of the medium, and we show that a well‐positioned image below the borehole can be obtained using interferometrically redatumed reflection responses with just a simple velocity model. We discuss some of the practical hurdles that restrict the application of the proposed method offshore.     

Two‐dimensional dictionary learning for suppressing random seismic noise

Dictionary learning is a successful method for random seismic noise attenuation that has been proven by some scholars. Dictionary learning–based techniques aim to learn a set of common bases called dictionaries from given noised seismic data. Then, the denoising process will be performed by assuming a sparse representation on each small local patch of the seismic data over the learned dictionary. The local patches that are extracted from the seismic section are essentially two‐dimensional matrices. However, for the sake of simplicity, almost all of the existing dictionary learning methods just convert each two‐dimensional patch into a one‐dimensional vector. In doing this, the geometric structure information of the raw data will be revealed, leading to low capability in the reconstruction of seismic structures, such as faults and dip events. In this paper, we propose a two‐dimensional dictionary learning method for the seismic denoising problem. Unlike other dictionary learning–based methods, the proposed method represents the two‐dimensional patches directly to avoid the conversion process, and thus reserves the important structure information for a better reconstruction. Our method first learns a two‐dimensional dictionary from the noisy seismic patches. Then, we use the two‐dimensional dictionary to sparsely represent all of the noisy two‐dimensional patches to obtain clean patches. Finally, the clean patches are patched back to generate a denoised seismic section. The proposed method is compared with the other three denoising methods, including FX‐decon, curvelet and one‐dimensional learning method. The results demonstrate that our method has better denoising performance in terms of signal‐to‐noise ratio, fault and amplitude preservation.     

Full waveform inversion of SH‐ and Love‐wave data in near‐surface prospecting

We develop a two‐dimensional full waveform inversion approach for the simultaneous determination of S‐wave velocity and density models from SH ‐ and Love‐wave data. We illustrate the advantages of the SH/Love full waveform inversion with a simple synthetic example and demonstrate the method's applicability to a near‐surface dataset, recorded in the village ?achtice in Northwestern Slovakia. Goal of the survey was to map remains of historical building foundations in a highly heterogeneous subsurface. The seismic survey comprises two parallel SH‐profiles with maximum offsets of 24 m and covers a frequency range from 5 Hz to 80 Hz with high signal‐to‐noise ratio well suited for full waveform inversion. Using the Wiechert–Herglotz method, we determined a one‐dimensional gradient velocity model as a starting model for full waveform inversion. The two‐dimensional waveform inversion approach uses the global correlation norm as objective function in combination with a sequential inversion of low‐pass filtered field data. This mitigates the non‐linearity of the multi‐parameter inverse problem. Test computations show that the influence of visco‐elastic effects on the waveform inversion result is rather small. Further tests using a mono‐parameter shear modulus inversion reveal that the inversion of the density model has no significant impact on the final data fit. The final full waveform inversion S‐wave velocity and density models show a prominent low‐velocity weathering layer. Below this layer, the subsurface is highly heterogeneous. Minimum anomaly sizes correspond to approximately half of the dominant Love‐wavelength. The results demonstrate the ability of two‐dimensional SH waveform inversion to image shallow small‐scale soil structure. However, they do not show any evidence of foundation walls.